Geology Reference
In-Depth Information
the pack ice moved at 10 km/day on average and revealed
one of the largest wind‐driven coastal polynyas.
Based on wavelet transform technique, another
method for ice tracking was developed at NASA/GSFC
[
Liu and Cavalieri
, 1998]. Wavelet transforms are analo-
gous to Fourier transforms but localized in both frequency
and time. The method is considered to be complementary
to the MCC as it improves the motion retrieval during
spring through late summer. It requires some prior knowl-
edge of the scale of ice motion, but it is computationally
efficient.
Zhao et al
. [2002] used this method to show
that the Arctic sea ice motion derived from SSM/I,
AMSR‐E, NSCAT, and QuikSCAT data were compatible
with each other and agreed well with ice motion derived
from buoy data. They also merged sea ice motion maps
obtained by averaging the weighted sea ice velocities
derived from QuikSCAT, SSM/I, AMSR‐E, and buoy
data to produce smoother sea ice motion with more com-
plete coverage. The merged Arctic sea ice motion data
during winter seasons from December 1988 to March
2003 were used to study the circulation regimes and win-
ter‐to‐winter variability of Arctic sea ice motion. Merging
ice motion derived from a few single‐channel data is an
approach that has been pursued in a few studies.
Radar images from SARs and scatterometers have been
used to track ice motion. Scatterometers are preferred for
global mapping because of their wide coverage. For
example, the Ku‐band NASA's NSCAT (operated from
September 1996 through June 1997) had two 600 km wide
swaths separated by a 200 km wide gap, with nominal
resolution of approximately 25 km.
Long and Drinkwater
[1999] used data from this scatterometer with the wavelet
analysis to track the ice motion in the Greenland Sea and
the Southern Ocean. One interesting conclusion from this
study was that the Antarctic ice might be correctly tracked
even during the austral spring melt onset when ERS C‐
band scatterometer data suffers from snow melting
effects. Once again, the study confirmed that ice motion
from NSCAT and SSM/I complement each other. The
merging of results between these two sensors in addition
to buoy data produced better composite maps.
A few researchers used the data from the next NASA's
Ku‐band scatterometer, Seawinds onboard QuikSCAT to
track ice motion in the Arctic. The sensor was launched
on 19 June 1999 and the mission ended on 19 November
2009. Ice motion field from a 12.5 km grid spacing from
QuikSCAT proved to be slightly more accurate than the
one from SSM/I [
Zhao et al
., 2002]. One advantage of
this sensor was its dual co‐polarization channels (HH
and VV). Each channel operated with a different scan-
ning geometry; the inner scan HH and the outer scan
VV had incidence angles of 46° and 54°, respectively.
This made the backscatter from the two channels essen-
tially independent and therefore allowed retrieval of
two independent ice drift (or motion) fields. A positive
consequence is the ability of filtering the final results
from the two channels for erroneous vectors by conducting
a comparison. [
Haarpaintner
[2006] used QuikSCATdata
with this approach to generate maps of sea ice drift in the
Arctic. He applied the MCC technique to the enhanced‐
resolution version of the data. The enhanced‐resolution
product consists of 36 h composites with 2.224 km grid
spacing, down from the original 12.5 km resolution [
Long
et al
. [1993]. Compared to the usually available 12.5 km
SSM/I 85.5 GHz data,
Haarpaintner
[2006] found that
ice motion from the enhanced-resolution QuikScat data
allowed detection of five times smaller speeds and
achieved an improvement of at least 25% in spatial accu-
racy. Figure 10.47 is an example of the 48 h displacement
field over the entire Arctic from 9-11 January, 2003.
White flags are tracking results at feasible nodes and
black flags are interpolated ice displacements after filter-
ing. Each drift vector is in good agreement with its neigh-
bors. A complete differential motion shows clearly the
Beaufort Gyre, the Transpolar Drift System, the Fram
Strait outflow, and the East Greenland current.
Arctic ice drift maps are produced regularly at the
French research Institute for Exploitation of the Sea
(acronym IFREMER following the name in French)
[
Girard‐Ardhuin and Ezraty
, 2012]. The maps are pro-
duced using the MCC technique applied to a suite of
satellite data, particularly from radiometers and scatter-
ometers. The sensors used include SSM/I, AMSR‐E,
QuikSAT, and Advanced Scatterometer (ASCAT). Drift
fields from single sensors at the same resolution are
combined into merged field maps built at 3 and 6 day
lags with a 62.5 km resolution (the spacing between the
inferred vectors). Fine resolution maps are generated
when necessary for a limited time in certain areas using
AVHRR or SAR. Maps are generated from September
to May. The root mean square drift difference varies
between 0.04 and 0.42 km/h depending on the sensor
and the season.
The displacement vector calculated using the MCC tech-
nique is subject to two sources of error [
Lindsay and Stern
,
2003]. The first is due to possible deviation in the geoloca-
tion of either one or both images. The second, called track-
ing error, is due to ambiguity in determining the location of
the ice in the second image that corresponds to the feature
(or grid point) in the first image, i.e., in determining the
location of the maximum correlation from equation
(10.108). This is more likely to happen when the feature
that is being tracked is not “unique” enough or if the
images have poor contrast. For that reason the images may
need contrast enhancement. Another reason for the ambi-
guity is the rotation of the ice, which is not taken into con-
sideration by the MCC technique. It can also be caused by
severe deformation in the ice between the dates of the two
